Re: [PATCH] usb: renesas_usbhs: fix __le16 warnings

[Simon Horman]

On Thu, Oct 17, 2019 at 08:57:26AM +0200, Geert Uytterhoeven wrote:
> Hi Shimoda-san, Simon,
> 
> On Thu, Oct 17, 2019 at 4:18 AM Yoshihiro Shimoda
> <yoshihiro.shimoda.uh@xxxxxxxxxxx> wrote:
> > > From: Simon Horman, Sent: Wednesday, October 16, 2019 9:27 PM
> > <snip>
> > > > diff --git a/drivers/usb/renesas_usbhs/common.c b/drivers/usb/renesas_usbhs/common.c
> > > > index 4c3de777ef6c..a3c30b609433 100644
> > > > --- a/drivers/usb/renesas_usbhs/common.c
> > > > +++ b/drivers/usb/renesas_usbhs/common.c
> > > > @@ -162,17 +162,17 @@ void usbhs_usbreq_get_val(struct usbhs_priv *priv, struct usb_ctrlrequest *req)
> > > >     req->bRequest           = (val >> 8) & 0xFF;
> > > >     req->bRequestType       = (val >> 0) & 0xFF;
> > > >
> > > > -   req->wValue     = usbhs_read(priv, USBVAL);
> > > > -   req->wIndex     = usbhs_read(priv, USBINDX);
> > > > -   req->wLength    = usbhs_read(priv, USBLENG);
> > > > +   req->wValue     = cpu_to_le16(usbhs_read(priv, USBVAL));
> > > > +   req->wIndex     = cpu_to_le16(usbhs_read(priv, USBINDX));
> > > > +   req->wLength    = cpu_to_le16(usbhs_read(priv, USBLENG));
> > >
> > > usbhs_read is backed by readl which performs
> > > a le->cpu conversion. Rather than have a double conversion
> > > perhaps it would be nicer to introduce usbhs_read_le.
> > > Likewise for write.
> >
> > I'm afraid but, I could not understand these comments.
> > At the moment, the usbhs_{read,write}() call io{read,write}16(),
> > not {read,write}l().
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/usb/renesas_usbhs/common.c?h=v5.4-rc3#n62
> 
> ioread16() and readw() don't do byteswapping on ARM, as ARM is
> little-endian. Likewise, cpu_to_le16() is a no-op on ARM.
> 
> Double swapping would matter only on a big-endian platform, and could
> indeed be avoided by introducing usbhs_read_le*() functions that are
> just wrappers around __raw_read16() on big-endian.
> However, until the Renesas USBHS IP core ends up on a big-endian
> platform, it's not worth doing that, IMHO.

Yes, that is all true.
I'm fine with this patch as it is.

> 
> > > >  }
> > > >
> > > >  void usbhs_usbreq_set_val(struct usbhs_priv *priv, struct usb_ctrlrequest *req)
> > > >  {
> > > >     usbhs_write(priv, USBREQ,  (req->bRequest << 8) | req->bRequestType);
> > > > -   usbhs_write(priv, USBVAL,  req->wValue);
> > > > -   usbhs_write(priv, USBINDX, req->wIndex);
> > > > -   usbhs_write(priv, USBLENG, req->wLength);
> > > > +   usbhs_write(priv, USBVAL,  le16_to_cpu(req->wValue));
> > > > +   usbhs_write(priv, USBINDX, le16_to_cpu(req->wIndex));
> > > > +   usbhs_write(priv, USBLENG, le16_to_cpu(req->wLength));
> > > >
> > > >     usbhs_bset(priv, DCPCTR, SUREQ, SUREQ);
> > > >  }
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> -- 
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
>